def func2():
yesterday = (datetime.datetime.now() - datetime.timedelta(days = 1)).strftime("%Y-%m-%d")
df_openA=pd.read_csv('openA.csv')
last_date_openA = time.strftime("%Y-%m-%d",time.strptime(str(int(df_openA.iloc[-1,0])),"%Y%m%d"))
day_gap_openA = w.tdayscount(last_date_openA, yesterday).Data[0][0]-1
print 2
w_wsd_data={}
stock_ls=[]
for stock in universe:
if stock[0]=='1':
stock = stock + '.SZ'
else:
stock += '.SH'
stock_ls.append(stock)
w_wsd_data[stock] = w.wsd(stock, 'open', (w.tdaysoffset(1,last_date_openA).Data[0][0]).strftime("%Y-%m-%d"),yesterday)
for i in xrange(day_gap_openA):
this_day = w.tdaysoffset(i+1,last_date_openA)
ls = []
this_day_csv = (this_day.Data[0][0]).strftime("%Y%m%d")
for stock in stock_ls:
ls.append(w_wsd_data[stock].Data[0][i])
f=open('openA.csv','a')
f.write(this_day_csv)
for stock_data in ls:
f.write(",")
f.write(str(stock_data))
f.write("\n")
f.close()
print 'finish'
def func9():
yesterday = (datetime.datetime.now() - datetime.timedelta(days = 1)).strftime("%Y-%m-%d")
df_discount_ratio_M = pd.read_csv('discount_ratio_M.csv')
last_date_discount_ratio_M = time.strftime("%Y-%m-%d",time.strptime(str(int(df_discount_ratio_M.iloc[-1,0])),"%Y%m%d"))
day_gap_discount_ratio_M = w.tdayscount(last_date_discount_ratio_M, yesterday).Data[0][0]-1
w_wsd_data={}
print 9
stock_ls=[]
for stock in universe:
if stock[0]=='1':
stock = stock + '.SZ'
else:
stock = stock + '.SH'
mA_code = w.wsd(stock, 'fund_smfcode', yesterday).Data[0][0]
stock_ls.append(mA_code)
w_wsd_data[mA_code] = w.wsd(mA_code, 'anal_tdiscountratio', (w.tdaysoffset(1,last_date_discount_ratio_M).Data[0][0]).strftime("%Y-%m-%d"),yesterday)
for i in xrange(day_gap_discount_ratio_M):
this_day = w.tdaysoffset(i+1,last_date_discount_ratio_M)
ls = []
this_day_csv = (this_day.Data[0][0]).strftime("%Y%m%d")
for stock in stock_ls:
ls.append(w_wsd_data[stock].Data[0][i])
f=open('discount_ratio_M.csv','a')
f.write(this_day_csv)
for stock_data in ls:
f.write(",")
f.write(str(stock_data))
f.write("\n")
f.close()
print 'finish'
def func11():
yesterday = (datetime.datetime.now() - datetime.timedelta(days = 1)).strftime("%Y-%m-%d")
df_industry=pd.read_csv('industry.csv')
last_date_industry = time.strftime("%Y-%m-%d",time.strptime(str(int(df_industry.iloc[-1,0])),"%Y%m%d"))
day_gap_industry = w.tdayscount(last_date_industry, yesterday).Data[0][0]-1
print 11
w_wsd_data={}
stock_ls=[]
for stock in list(df_industry.columns.values)[1:] :
stock_ls.append(stock)
w_wsd_data[stock] = w.wsd(stock, 'close', (w.tdaysoffset(1,last_date_industry).Data[0][0]).strftime("%Y-%m-%d"),yesterday)
print w_wsd_data
f=open('industry.csv','a')
for i in xrange(day_gap_industry):
this_day = w.tdaysoffset(i+1,last_date_industry)
ls = []
this_day_csv = (this_day.Data[0][0]).strftime("%Y%m%d")
for stock in stock_ls:
ls.append(w_wsd_data[stock].Data[0][i])
f.write(this_day_csv)
for stock_data in ls:
f.write(",")
f.write(str(stock_data))
f.write("\n")
f.close()
print 'finish'
def get_trading_date_from_now(date_now, diff_periods, period=ql.Days):
always_using_ql = False # 是否全部使用quantlib处理日期，否则只在某些情形使用quantlib处理日期
if (int(date_now.split('-')[0]) <= 2019 and int(date_now.split('-')[1]) <= 9 and diff_periods <= 1) or (int(date_now.split('-')[0]) <= 2018 and diff_periods <= 1) or always_using_ql:
calculation_date = ql.Date(int(date_now.split('-')[2]), int(date_now.split('-')[1]),
int(date_now.split('-')[0]))
calendar = ql.China()
date_diff = calendar.advance(calculation_date, diff_periods, period).to_date().strftime('%Y-%m-%d')
else: # 其余日子用wind处理
if period==ql.Days:
w.start()
date_diff = w.tdaysoffset(diff_periods, date_now, '').Data[0][0].strftime('%Y-%m-%d')
if diff_periods == 0: # 非交易日顺延到下一个交易日
date_now_date = datetime.date.fromisoformat(date_now)
date_diff_date = datetime.date.fromisoformat(date_diff)
if date_diff_date < date_now_date:
date_diff = w.tdaysoffset(1, date_now, '').Data[0][0].strftime('%Y-%m-%d')
elif period==ql.Months:
w.start()
date_diff = w.tdaysoffset(diff_periods, date_now, 'Period=M').Data[0][0].strftime('%Y-%m-%d')
if diff_periods == 0: # 非交易日顺延到下一个交易日
date_now_date = datetime.date.fromisoformat(date_now)
date_diff_date = datetime.date.fromisoformat(date_diff)
if date_diff_date < date_now_date:
date_diff = w.tdaysoffset(1, date_now, '').Data[0][0].strftime('%Y-%m-%d')
return date_diff
def get_trading_date_from_now(date_now, diff_periods, period=ql.Days):
if period == ql.Days:
w.start()
date_diff = w.tdaysoffset(diff_periods, date_now,
'').Data[0][0].strftime('%Y-%m-%d')
if diff_periods == 0:
date_now_date = datetime.date.fromisoformat(date_now)
date_diff_date = datetime.date.fromisoformat(date_diff)
if date_diff_date < date_now_date:
date_diff = w.tdaysoffset(1, date_now,
'').Data[0][0].strftime('%Y-%m-%d')
return date_diff
# 未改善代码
calculation_date = ql.Date(int(date_now.split('-')[2]),
int(date_now.split('-')[1]),
int(date_now.split('-')[0]))
calendar = ql.China()
date_diff = calendar.advance(calculation_date, diff_periods,
period).to_date().strftime('%Y-%m-%d')
# 临时修正错误，是包QuantLib的错误，节假日问题
if date_now == '2019-01-02' and diff_periods == -1 and period == ql.Days:
return '2018-12-28'
if (date_now == '2019-02-08' or date_now == '2019-02-09'
or date_now == '2019-02-10' or date_now
== '2019-02-11') and diff_periods == -1 and period == ql.Days:
return '2019-02-01'
if (date_now == '2019-05-06' or date_now == '2019-05-02'
or date_now == '2019-05-03' or date_now
== '2019-05-04') and diff_periods == -1 and period == ql.Days:
return '2019-04-30'
return date_diff
def tradedate(days=30):
lastdate = datetime.now().date()
starttradedate = (datetime.now() - timedelta(days)).date()
if lastdate.weekday() == 5 or lastdate.weekday() == 6:
lasttradedate = w.tdaysoffset(0, lastdate).Data[0][0].date()
else:
lasttradedate = w.tdaysoffset(-1, lastdate).Data[0][0].date()
if starttradedate >= lasttradedate:
starttradedate = lasttradedate
else:
pass
return (str(starttradedate), str(lasttradedate))
def lookup_w(w_panel, t, calendar, pre_w=None, pre_r=None):
"""
用于返回对应高频（这里是daily）日期在回测期间的权重（调频周期为低频周、月）.
:param w_panel: dict，在调仓日（收盘后）的对应新权重.
:param t: str, 需要确定组合日收益率的当日.
:param pre_w: list, 前2交易日收盘后的权重，即是前1交易日开盘前权重.
:param pre_r: list, 前1交易日当天收益率.
:return: list,t当日开盘前权重.
"""
assert isinstance(w_panel, dict), "TypeError: w_panel should be type dict"
# 按时间顺序排序
w_panel = OrderedDict(sorted(w_panel.items(), key=lambda k: k[0]))
rebalance_seq = list(w_panel.keys())
# 由上期莫转换到本期初
try:
tt = rebalance_seq[0]
rebalance_seq_ = list(map(lambda tt: w.tdaysoffset(1, tt, f'''TradingCalendar={
calendar}''').Data[0][0].strftime(
"%Y-%m-%d"),
rebalance_seq))
except AttributeError as e:
print(f"{tt} {e}")
if t[:10] in rebalance_seq_:
idx = rebalance_seq_.index(t[:10])
cur_w = w_panel[rebalance_seq[idx]]
else:
# 价格变化导致每天的实际权重发生变化
cur_w = list(np.array(pre_w) * (np.array(pre_r)/100 + 1) / np.dot(np.array(pre_r)/100 + 1,
np.array(pre_w)))
return cur_w
def get_schedule(**kwargs):
sec_code = kwargs.get('sec_code')
end_date = kwargs.get('end_date')
if not sec_code or not end_date:
raise ValueError('sec_code and end_date should not be empty')
participant_rate = kwargs.get('participant_rate') or 0.15
target_ratio = kwargs.get('target_ratio') or 0.01
period = kwargs.get('period') or 100
target_vol = kwargs.get('target_vol') or 8000000
target_period = kwargs.get('target_period')
price_ratio = kwargs.get('price_ratio') or 0.95
update = kwargs.get('update') or False
ret_vol, ret_price = get_trade_schedule(participant_rate=participant_rate,
target_ratio=target_ratio,
sec_code=sec_code,
end_date=end_date,
period=period,
target_vol=target_vol,
target_period=target_period,
price_ratio=price_ratio)
# FIXME fix the border
print(sum(ret_vol), target_vol)
next_date = w.tdaysoffset(1, end_date).Data[0][0].strftime('%Y-%m-%d')
sec_code = sec_code
vol = ret_vol[0]
price = min(ret_price)
df = pd.DataFrame([[sec_code, next_date, vol, price]],
columns=['sec_code', 'trade_date', 'vol', 'price'])
curr_df = pd.read_csv('data/trade_strategy.csv')
curr_df = curr_df.append(df)
if update:
curr_df.to_csv('data/trade_strategy.csv', index=False)
return df
def GetLastTradeDate():
end_date = date.today()
count = w.tdayscount(end_date)
if count.Data == 0 or datetime.now().hour <= 15:
return w.tdaysoffset(-1, end_date).Data[0][0].date()
return end_date
def histroy_same_date(tdate, years=10):
his_list = []
for y in range(-years, 0):
his_y = w.tdaysoffset(y, "2018-04-24", "Period=Y;TradingCalendar=DCE")
his_y = his_y.Data[0][0]
his_list.append(his_y)
return his_list
def post(self):
"""
json str:{"offset": "1", "begin_time": "2017-3-31", "options": ""}
:return: 返回万得返回数据dict
"""
data_dic = request.json
logger.info('/tdaysoffset/ data_dic:%s', data_dic)
offset = int(data_dic['offset'])
begin_time = data_dic['begin_time']
options = data_dic['options']
if not w.isconnected():
w.start()
if options == "":
options = None
ret_data = w.tdaysoffset(offset, begin_time, options)
error_code = ret_data.ErrorCode
if error_code != 0:
msg = ERROR_CODE_MSG_DIC.setdefault(error_code, "")
logger.error('tdaysoffset("%s", "%s", "%s") ErrorCode=%d %s' %
(offset, begin_time, options, error_code, msg))
return {'error_code': ret_data.ErrorCode, 'message': msg}, 404
# if ret_data.ErrorCode != 0:
# logger.error(
# 'tdaysoffset("%s", "%s", "%s") ErrorCode=%d' % (offset, begin_time, options, ret_data.ErrorCode))
# return {'error_code': ret_data.ErrorCode}, 404
# 将 Data数据中所有 datetime date 类型的数据转换为 string
if len(ret_data.Data) > 0 and len(ret_data.Data[0]) > 0:
date_str = format_2_date_str(ret_data.Data[0][0])
else:
logger.warning('tdaysoffset(%s, %s, %s) No value return' %
(offset, begin_time, options))
date_str = ''
ret_dic = {'Date': date_str}
# print('offset:\n', ret_dic)
return ret_dic
def t_days_offset(offset=0, cur_date=datetime.now()):
try:
w.start()
result = w.tdaysoffset(offset, cur_date, "").Data[0][0]
return result
except IndexError as e:
print(format(e))
raise
def advance_date(date, unit, freq):
try:
ret = w.tdaysoffset(int(unit) * -1, date, 'period=' + freq)
WindDataProvider.force_throw_err(ret,
'WindDataProvider.advance_date')
return ret.Data[0][0]
except NameError:
pass
def get_trading_date_from_now(date_now, diff_periods, period=ql.Days):
if period==ql.Months:
w.start()
date_diff = w.tdaysoffset(diff_periods, date_now, 'Period=M').Data[0][0].strftime('%Y-%m-%d')
if diff_periods == 0: # 非交易日顺延到下一个交易日
date_now_date = datetime.date.fromisoformat(date_now)
date_diff_date = datetime.date.fromisoformat(date_diff)
if date_diff_date < date_now_date:
date_diff = w.tdaysoffset(1, date_now, '').Data[0][0].strftime('%Y-%m-%d')
else: # 默认以日为单位进行调整
w.start()
date_diff = w.tdaysoffset(diff_periods, date_now, '').Data[0][0].strftime('%Y-%m-%d')
if diff_periods == 0: # 非交易日顺延到下一个交易日
date_now_date = datetime.date.fromisoformat(date_now)
date_diff_date = datetime.date.fromisoformat(date_diff)
if date_diff_date < date_now_date:
date_diff = w.tdaysoffset(1, date_now, '').Data[0][0].strftime('%Y-%m-%d')
return date_diff
def _get_trade_day(self, caldate):
res = w.tdaysoffset(
0,
caldate,
)
if res.ErrorCode != 0:
raise WindQueryFailError(
"Get date data failed, errorcode={}.".format(res.ErrorCode))
return res.Data[0][0]
def get_stdevr(stocks, date):
date_i = date.strftime("%Y%m%d")
date_20 = w.tdaysoffset(-20, date, "").Data[0][0].strftime("%Y%m%d")
date_60 = w.tdaysoffset(-60, date, "").Data[0][0].strftime("%Y%m%d")
rawdata1 = w.wss(
stocks, "stdevr", "startDate=" + date_20 + ";endDate=" + date_i +
";period=1;returnType=1")
data1 = pd.DataFrame(rawdata1.Data).T
rawdata2 = w.wss(
stocks, "stdevr", "startDate=" + date_60 + ";endDate=" + date_i +
";period=1;returnType=1")
data2 = pd.DataFrame(rawdata2.Data).T
data = pd.concat([data1, data2], axis=1)
data.index = stocks
data.columns = ['std20', 'std60']
return data
def get_all_month_start_end_dates(start_date='', end_date=''):
'''
format for start_date and end_date: 'yyyymmdd'
:param start_date:
:param end_date:
:return: list of trading dates, with format 'yyyydddd'
'''
# FIXME check the missing trade cal table
# rows, cols = get_dates_statics(start_date, end_date)
# return list(set([(item[6].strftime('%Y%m%d'), item[7].strftime('%Y%m%d')) for item in rows]))
from WindPy import w
w.start()
_start_date = start_date[:6] + '01'
_start_date = w.tdaysoffset(-1, _start_date).Data[0][0]
data = w.tdays(_start_date, end_date, "Period=M").Data[0]
ret = []
for item in data:
_next_start = w.tdaysoffset(1, item).Data[0][0]
ret.append(item.strftime('%Y%m%d'))
ret.append(_next_start.strftime('%Y%m%d'))
return ret[1:-1]
def _eval_seq_d(self, seq_d):
if not w.isconnected():
w.start()
sleep(3)
seq = seq_d.copy()
seq.pop(0)
last = seq[-1]
next_ = w.tdaysoffset(
1, last,
f"TradingCalendar={self.calendar}").Data[0][0].strftime("%Y-%m-%d")
seq.append(next_)
return seq
def func6():
yesterday = (datetime.datetime.now() - datetime.timedelta(days = 1)).strftime("%Y-%m-%d")
df_closeB=pd.read_csv('closeB.csv')
last_date_closeB = time.strftime("%Y-%m-%d",time.strptime(str(int(df_closeB.iloc[-1,0])),"%Y%m%d"))
day_gap_closeB = w.tdayscount(last_date_closeB, yesterday).Data[0][0]-1
print 6
w_wsd_data={}
stock_ls=[]
for stock in universe:
if stock =='150073':
stock='150075.SZ'
else:
if stock[0]=='1':
stock = str(int(stock[:6])+1) + '.SZ'
else:
stock = str(int(stock[:6])+1) + '.SH'
stock_ls.append(stock)
w_wsd_data[stock] = w.wsd(stock, 'close', (w.tdaysoffset(1,last_date_closeB).Data[0][0]).strftime("%Y-%m-%d"),yesterday)
# print 1
print stock
#if w_wsd_data[stock].Data[0][0]==None:
for i in xrange(day_gap_closeB):
this_day = w.tdaysoffset(i+1,last_date_closeB)
ls = []
this_day_csv = (this_day.Data[0][0]).strftime("%Y%m%d")
for stock in stock_ls:
ls.append(w_wsd_data[stock].Data[0][i])
f=open('closeB.csv','a')
f.write(this_day_csv)
for stock_data in ls:
f.write(",")
f.write(str(stock_data))
f.write("\n")
f.close()
print 'finish'
def get_nearest_hrange(now, num_h, calendar):
"""
根据当天需要确定下期持仓权重的时间，以及优化需要的历史K线个数，确定提取数据的开始和结束时间.
:param now: str,当前时间.
:param num_h: int,估计均值、方差需要的历史K线数.
:return: str,str;开始时间，结束时间
"""
dd, hh = now.split(" ")
ft = max(filter(lambda x: x < hh, ["00"]+HOUR_BAR))
if ft == "00":
ft = HOUR_BAR[-1]
dd_ = w.tdaysoffset(-1, dd, f"TradingCalendar={calendar}").Data[0][0].strftime("%Y-%m-%d")
idx_t = 0
else:
idx_t = HOUR_BAR.index(ft) + 1
dd_ = dd
days, idx_h = divmod(num_h-idx_t, 4)
fh = HOUR_BAR[-idx_h]
ddh = w.tdaysoffset(-days-int(idx_h > 0), dd, f"TradingCalendar={calendar}").Data[0][0].strftime("%Y-%m-%d")
return " ".join([ddh, fh]), " ".join([dd_, ft])
def get_data(now_date):
w.start()
end_date = w.tdaysoffset(-1, now_date, "").Data[0][0].strftime('%Y-%m-%d')
start_date = w.tdaysoffset(-241, now_date,
"").Data[0][0].strftime('%Y-%m-%d')
file_path = 'data\\in_trading\\data.csv'
if os.path.exists(file_path): # 存在数据文件
df = pd.read_csv(file_path, index_col=0)
index = df.index.values
if index[0] != start_date or index[-1] != end_date: # 时间日期不符
os.remove(file_path) # 删除数据文件
data = w.wsd(SW1, "pct_chg", start_date, end_date, "PriceAdj=T")
df = pd.DataFrame(data=np.array(data.Data).transpose(),
index=data.Times,
columns=SW1.split(','))
df.to_csv(file_path)
else: # 不存在数据文件
data = w.wsd(SW1, "pct_chg", start_date, end_date, "PriceAdj=T")
df = pd.DataFrame(data=np.array(data.Data).transpose(),
index=data.Times,
columns=SW1.split(','))
df.to_csv(file_path)
def _calculate_factor(self):
date_list = self.date
startDate = str(
w.tdaysoffset(-self.N, "".join(date_list), "").Data[0][0]) # 区间数据
vol_data = np.array(
w.wss(
self.code_list, "avg_turn_per", "startDate=" +
"".join(startDate) + ";endDate=" + "".join(date_list)).Data[0])
vol_data = vol_data / 100.0
VOL240 = pd.DataFrame(data=vol_data,
index=self.code_list,
columns=[self.factor_name])
return VOL240
def get_history(date_end, T):
w.start()
print('获取指数历史数据......')
if type(T) == int:
date_start = w.tdaysoffset(-T, date_end,
'').Data[0][0].strftime('%Y-%m-%d')
else:
date_start = T
Data = w.wsd(list(SW1.keys()), "pct_chg", date_start, date_end, "")
return_d = np.array(Data.Data)
date_list = Data.Times
date_list = [d.strftime('%Y-%m-%d') for d in date_list]
print('指数历史数据获取完毕')
return return_d, date_list
def update_recorder(endDate=None, recordFile=None, holdListPath=None):
endDate = dt.datetime.today().strftime(
'%Y%m%d') if endDate is None else endDate
recordFile = r'.\backtest_recorder.csv' if recordFile is None else recordFile
holdListPath = r'.\holdLists' if holdListPath is None else holdListPath
recordDF = pd.read_csv(recordFile)
lastUpdt = recordDF['tradeDate'].values[-1]
w.start()
startDate = w.tdaysoffset(1, str(lastUpdt)).Data[0][0].strftime('%Y%m%d')
if startDate > endDate:
print('no new data to update')
return
betweenDays = w.tdays(startDate, endDate).Data[0]
betweenDays = [int(tdt.strftime('%Y%m%d')) for tdt in betweenDays]
print(betweenDays)
outPut = {
'tradeDate': [],
'buyReturn': [],
'holdReturn': [],
'netReturn': []
}
cols = ['stkcd', 'buyDate', 'ret', 'flag']
for tdt in betweenDays:
retList = pd.read_csv(os.path.join(
holdListPath, 'returns_tradeDate_{}.csv'.format(tdt)),
encoding='gbk')
retList = retList.loc[:, cols]
validBuy = np.all(
[retList['buyDate'] == tdt,
np.isin(retList['flag'], (0, 3, 4))],
axis=0)
validHold = np.all(
[retList['buyDate'] < tdt,
np.isin(retList['flag'], (0, 2, 3, 4))],
axis=0)
outPut['tradeDate'].append(tdt)
outPut['buyReturn'].append(retList.loc[validBuy, 'ret'].mean())
outPut['holdReturn'].append(retList.loc[validHold, 'ret'].mean())
outPut['netReturn'].append(retList.loc[validBuy | validHold,
'ret'].mean())
outPut = pd.DataFrame(
outPut).loc[:, ['tradeDate', 'buyReturn', 'holdReturn', 'netReturn']]
indices = ['000001.SH', '000300.SH', '000905.SH', '000016.SH']
idxRets = pd.DataFrame(w.wsd(','.join(indices), 'pct_chg', startDate,
endDate).Data,
columns=indices)
idxRets = idxRets.loc[:] / 100
outPut = pd.concat([outPut, idxRets], axis=1)
outPut.to_csv(recordFile, mode='a+', index=False, header=False)
print('recorder updated till {}'.format(endDate))
def _calculate_factor(self):
date_list = self.date
startDate = str(
w.tdaysoffset(-self.length, "".join(date_list),
"").Data[0][0]) # 每个月22个交易日
beta_data = np.array(
w.wss(
self.code_list, "beta",
"startDate=" + startDate + ";endDate=" + "".join(date_list) +
";period=2;returnType=1;index=" + self.refer_index).Data[0])
BETA = pd.DataFrame(data=beta_data,
index=self.code_list,
columns=[self.factor_name])
return BETA
def get_tdays_offset(self, begin_date, offset, cycle="", style=""):
"""
获取偏移日期
输入参数：
begin_date 开始日期 yyyymmdd 或 yyyy-mm-dd
offset 偏移数量 -5表示往前5个日期；5表示往后5个日期
cycle和style同get_tdays函数中
函数返回：
返回偏移N个周期的日期
类型：yyyymmdd
"""
option = {"Days": style, "Period": cycle}
td = w.tdaysoffset(offset, begin_date, **option)
return td.Times[0].strftime("%Y%m%d")
def train_test_model(vol_df, short_term_days, features, model, train_date,
test_date):
predX = vol_df.ix[-short_term_days - 1:, features]
pred_end_date = w.tdaysoffset(short_term_days, predX.index[-1]).Data[0][0]
pred_dates = w.tdays(predX.index[-1].strftime("%Y-%m-%d"),
pred_end_date.strftime("%Y-%m-%d")).Times
predX.index = pred_dates
vol_df[features] = vol_df[features].shift(short_term_days)
vol_df.dropna(inplace=True)
train_df = vol_df[vol_df.index <= test_date]
test_df = vol_df[vol_df.index >= test_date]
train_df = train_df.append(test_df.ix[0]) # 为了让图看上去连续
trainX, trainY = train_df[features], train_df['vol']
testX, testY = test_df[features], test_df['vol']
# 用Lasso来做feature selection
lasso = linear_model.Lasso(alpha=0.0005)
lasso.fit(trainX, trainY)
sfm = SelectFromModel(lasso, prefit=True)
fea_trainX = sfm.transform(trainX)
fea_testX = sfm.transform(testX)
fea_predX = sfm.transform(predX)
if model == "LinearRegression":
reg_model = linear_model.LinearRegression()
elif model == "KernelRidgeRegression":
reg_model = KernelRidge(kernel='rbf')
elif model == "SupportVectorRegression":
reg_model = SVR(kernel="linear")
elif model == "Ridge":
reg_model = linear_model.Ridge()
elif model == "RandomForestRegression":
reg_model = RandomForestRegressor()
elif model == "AdaBoostRegression":
reg_model = AdaBoostRegressor()
else:
return NotImplementedError
reg_model.fit(fea_trainX, trainY)
train_pred = reg_model.predict(fea_trainX)
test_pred = reg_model.predict(fea_testX)
pred_pred = reg_model.predict(fea_predX)
return (pd.DataFrame({'pred': train_pred}, index=trainX.index),
pd.DataFrame({'pred': test_pred}, index=testX.index),
pd.DataFrame({'pred': pred_pred}, index=predX.index))
def generate_return_report(tradeDate, tradeListPath=None, holdListPath=None):
tradeDate = str(tradeDate)
holdListPath = r'.\holdLists' if holdListPath is None else holdListPath
tradeListPath = r'.\buyLists' if tradeListPath is None else tradeListPath
w.start()
infoDate = w.tdaysoffset(-1, tradeDate).Data[0][0].strftime('%Y%m%d')
retList = pd.read_csv(os.path.join(
holdListPath, 'returns_tradeDate_{}.csv'.format(infoDate)),
encoding='gbk')
trdList = pd.read_csv(os.path.join(
tradeListPath, 'stkNum50_tradeList_infoDate_{}.csv'.format(infoDate)),
encoding='gbk')
holdStkList = retList.loc[retList['buyDate'] == int(infoDate),
['stkcd', 'predictVal']].set_index(['stkcd'])
buyStkList = trdList.loc[:, ['stkcd', 'predictVal']].set_index(['stkcd'])
conn = cx_Oracle.connect(r'c##sensegain/[email protected]/wind')
cursor = conn.cursor()
sqlLines = 'SELECT TRADE_DT, S_INFO_WINDCODE, S_DQ_OPEN, S_DQ_CLOSE, S_DQ_PRECLOSE, S_DQ_PCTCHANGE/100, S_DQ_CLOSE/S_DQ_OPEN-1, S_DQ_TRADESTATUS ' \
'FROM c##wind.AShareEODPrices WHERE TRADE_DT={0}'.format(tradeDate)
rets = cursor.execute(sqlLines).fetchall()
rets = pd.DataFrame(rets,
columns=[
'tradeDate', 'stkcd', 'open', 'close', 'preclose',
'retCC', 'retOC', 'trdStat'
])
rets['flagOC'] = 1 * (rets['trdStat'] == '停牌') + 2 * (
rets['open'] / rets['preclose'] >=
1 + 0.099) + 3 * (rets['open'] / rets['preclose'] <= 1 - 0.099)
rets['flagCC'] = 1 * (rets['trdStat'] == '停牌') + 2 * (
rets['retCC'] >= 0.099) + 3 * (rets['retCC'] <= -0.099)
rets['stkcd'] = rets['stkcd'].map(lambda x: int(x[:6]))
rets.set_index(['stkcd'], inplace=True)
holdStkList = holdStkList.join(
rets.loc[:, ['tradeDate', 'close', 'retCC', 'flagCC']])
holdStkList.columns = ['predictVal', 'date', 'close', 'ret', 'flag']
holdStkList['buyDate'] = infoDate
buyStkList = buyStkList.join(
rets.loc[:, ['tradeDate', 'close', 'retOC', 'flagOC']])
buyStkList.columns = ['predictVal', 'date', 'close', 'ret', 'flag']
buyStkList['buyDate'] = tradeDate
output = pd.concat([holdStkList, buyStkList], axis=0)
output = output.loc[:, ['date', 'close', 'ret', 'flag', 'buyDate']]
output.to_csv(
os.path.join(holdListPath,
'returns_tradeDate_{}.csv'.format(tradeDate)))
print('return report generated for trade date {}'.format(tradeDate))
def post(self):
"""
json str:{"offset": "1", "begin_time": "2017-3-31", "options": ""}
:return: 返回万得返回数据dict
"""
args = tdays_offset_parser.parse_args()
logger.info('/tdaysoffset/ args:%s', args)
# offset = int(args['offset'])
# begin_time = args['begin_time']
# options = args['options']
if args['options'] == "":
args['options'] = None
if not w.isconnected():
w.start()
ret_data = None
for nth in range(2):
ret_data = w.tdaysoffset(**args)
error_code = ret_data.ErrorCode
if error_code != 0:
if nth == 0 and error_code == -40521010:
w.stop()
w.start()
logger.warning('尝试重新登陆成功，再次调用函数')
continue
msg = ERROR_CODE_MSG_DIC.setdefault(error_code, "")
logger.error('tdaysoffset("%s") ErrorCode=%d %s' %
(args, error_code, msg))
raise RequestError(msg, None, error_code)
else:
break
else:
if ret_data is None:
msg = 'wst(%s) ret_data is None' % args
logger.error(msg)
raise RequestError(msg, None, 0)
# 将 Data数据中所有 datetime date 类型的数据转换为 string
if len(ret_data.Data) > 0 and len(ret_data.Data[0]) > 0:
date_str = format_2_date_str(ret_data.Data[0][0])
else:
logger.warning('tdaysoffset(%s) No value return' % args)
date_str = ''
ret_dic = {'Date': date_str}
# print('offset:\n', ret_dic)
return ret_dic
def get_backtest_first_date(self, num_d):
"""
返回第一次回测所需要提取数据的最早日期.
.. deprecated:: v.6.0.3
Use :func:`Mkv_data2.get_nearest_hrange` instead.
:param num_d: 估计组合收益率统计量所需K线个数.
:return: str, 开始时间.
"""
if not w.isconnected():
w.start()
sleep(3)
first = w.tdaysoffset(
-(num_d - 1), self._start,
f"TradingCalendar={self.calendar}").Data[0][0].strftime("%Y-%m-%d")
return first
def _calculate_factor(self):
date_list = self.date
startDate = str(
w.tdaysoffset(-self.N, "".join(date_list), "").Data[0][0]) # 区间数据
price_data = np.array(
w.wss(
self.code_list, "high_per,low_per,close_per",
"startDate=" + "".join(startDate) + ";endDate=" +
"".join(date_list) + ";priceAdj=T").Data)
avg_price = np.sum(price_data, axis=0) / 3.0
volume = np.array(
w.wss(
self.code_list, "vol_per", "unit=1;startDate=" +
"".join(startDate) + ";endDate=" + "".join(date_list)).Data[0])
result = avg_price * volume
MoneyFlow20 = pd.DataFrame(data=result,
index=self.code_list,
columns=[self.factor_name])
return MoneyFlow20
def eval_returns_drange(self,
codes,
eval_begin,
eval_end,
get_t_seq=False):
"""
获取"D"频率下，指定开始和结束日期内每个K线指定股票变量值.
:param codes: list,股票代码.
:param eval_begin: str, 开始日期.
:param eval_end: str, 结束日期.
:param get_t_seq: bool, 是否返回区间内日期.
:return: list or DataFrame, list; 变量对象,日期.
"""
if not w.isconnected():
w.start()
sleep(3)
assert self._freq == "D", "ParameterError: method eval_returns_drange仅支持D为频率"
# eval_begin = w.tdaysoffset(-1, eval_begin,
# f"TradingCalendar={self.calendar}").Data[0][0].strftime(
# "%Y-%m-%d")
# res = w.wsd(codes, "close", eval_begin, eval_end,
# f"showblank=0;PriceAdj=F;TradingCalendar={self.calendar}")
# 转换为工作日
eval_begin = w.tdaysoffset(
-1, eval_begin, "Days=Weekdays").Data[0][0].strftime("%Y-%m-%d")
res = w.wsd(codes, "close", eval_begin, eval_end,
"showblank=0;PriceAdj=F;Days=Weekdays")
res.Data = close2return(res.Data,
percentage=True,
tolist=True,
fillna_value=0.0)
returns = res.Data
res.Times = res.Times[1:]
cash_days = {"M": 30, "W": 7, "D": 1}
returns.append([self._cash_r * cash_days[self._freq]] *
len(returns[0]))
returns = list(zip(*returns))
if get_t_seq:
returns = returns, list(
map(lambda x: x.strftime("%Y-%m-%d"), res.Times))
return returns
def date_offset(day,sft=1,Days='T',Period='',MKT='A', tp=''):
'''
wind get date
:day:
:sft: shift
:Days: W weekday C calendar T tradeday
:Period: W M Q S Y
:MKT: HK A
'''
from WindPy import w
w.start()
DaysD = {'W': 'Weekdays', 'C':'Alldays'}
MarketD = {'HK': 'HKEX', 'A':''}
suffix = ''
if Days!='T': suffix += 'Days=%s;'%DaysD[Days]
if Period!='': suffix += 'Period=%s;'%Period
if MKT!='A': suffix += 'MKT=%s;'%MarketD[MKT]
day0 = w.tdaysoffset(sft, day, suffix).Data[0][0]
return __dtTrans(day0, tp)
def get_tdaysoffset(num, date, **options):
'''
获取日期端点数据。
Parameters
-----------
num
int,正为面向未来,负为面向过去
date
str,'2018-05-10'
Returns
-------
datetime
'''
options = dict_2_str(options)
date = w.tdaysoffset(num, date, options)
return date.Data[0][0]
__author__ = 'aming.tao'
from WindPy import w
from datetime import *
w.start()
data=w.wsd("600000.SH","close,amt","2013-04-30", datetime.today()-timedelta(1))#取浦发银行收盘价等信
data=w.wsd("600000.SH","close,amt", datetime.today()-timedelta(100))#
data=w.wsi("600000.SH","close,amt","2015-10-01 9:00:00")#取浦发银行分钟收盘价等信息
data=w.wst("600000.SH","open", datetime.today()-timedelta(0,2*3600), datetime.now())#取浦发银行tick数据信息
data=w.wss("600000.SH,000001.SZ","eps_ttm,orps,surpluscapitalps","rptDate=20121231")#取浦发银行等财务数据信息
data=w.wset("SectorConstituent",u"date=20130608;sector=全部A股")#取全部A 股股票代码、名称信息
w.wset("IndexConstituent","date=20130608;windcode=000300.SH;field=wind_code,i_weight")#取沪深300 指数中股票代码和权重
w.wset("TradeSuspend","startdate=20130508;enddate=20130608;field=wind_code,sec_name,suspend_type,suspend_reason")#取停牌信息
w.wset("SectorConstituent",u"date=20130608;sector=风险警示股票;field=wind_code,sec_name")#取ST 股票等风险警示股票信息
w.tdays("2013-05-01","2013-06-08")#返回5 月1 日到6 月8 日之间的交易日序列
w.tdays("2013-05-01")#返回5 月1 日到当前时间的交易日序列
w.tdaysoffset(-5,"2013-05-01")#返回5 月1 日前推五个交易日的日期，返回2013-4-19
w.tdaysoffset(-5)#返回当前时间前推五个交易日的日期
w.tdayscount("2013-05-01","2013-06-08")#返回5 月1 日到6 月8 日之间的交易日序列长度，为27
w.stop()
def tdaysoffset(n, date):
tmp = w.tdaysoffset(n, date)
return tmp.Data[0][0]
